accelerated mitochondrial evolution and ... - genetics · speciation. therefore, a major focus of...
TRANSCRIPT
Copyright � 2008 by the Genetics Society of AmericaDOI: 10.1534/genetics.107.081364
Accelerated Mitochondrial Evolution and ‘‘Darwin’s Corollary’’: AsymmetricViability of Reciprocal F1 Hybrids in Centrarchid Fishes
Daniel I. Bolnick,*,1 Michael Turelli,† Hernan Lopez-Fernandez,‡ Peter C. Wainwright† andThomas J. Near§
*Section of Integrative Biology, University of Texas, Austin, Texas 78712, †Section of Evolution and Ecology, University of California, Davis,California 95616, ‡Department of Wildlife and Fisheries Sciences, Texas A&M University, College Station, Texas 77843-2258 and §Department
of Ecology and Evolution and Peabody Museum of Natural History, Yale University, New Haven, Connecticut 06520
Manuscript received August 29, 2007Accepted for publication November 26, 2007
ABSTRACT
Reciprocal crosses between species can yield hybrids with different viabilities. The high frequency of thisasymmetric hybrid viability (‘‘Darwin’s corollary’’) places it alongside Haldane’s rule and the ‘‘large-Xeffect’’ as a general feature of postmating reproductive isolation. Recent theory suggests that reciprocalcross asymmetries can arise from stochastic substitutions in uniparentally inherited loci such as mito-chondrial genomes, although large systematic differences in mitochondrial substitution rates can alsocontribute to asymmetries. Although the magnitude of asymmetry will be relatively insensitive to unequalrates of mitochondrial evolution in diverging species, we show here that rate asymmetries can have a largeeffect on the direction of viability asymmetries. In reciprocal crosses between species, the maternal parentwith faster mitochondrial evolution will tend to produce less viable F1 hybrids owing to an increasedprobability of mito-nuclear incompatibilities. We test this prediction using data on reciprocal hybridviability and molecular evolution rates from a clade of freshwater fishes, Centrarchidae. As predicted,species with accelerated mitochondrial evolution tend to be the worse maternal parent for F1 hybrids,providing the first comparative evidence for a systematic basis to Darwin’s corollary. This result is con-sistent with the hypothesis that mito-nuclear incompatibilities can play an important role in reproductiveisolation. Such asymmetrical reproductive isolation may help explain the asymmetrical mitochondrialintrogression observed between many hybridizing species. However, as with any comparative study, wecannot rule out the possibility that our results arise from a mutual correlation with a third variable such asbody size.
WHEN species hybridize, their progeny often ex-hibit some degree of infertility or inviability. This
hybrid breakdown prevents introgression between theparental species and so can play an important role inspeciation. Therefore, a major focus of evolutionarygenetics research has been to understand the geneticbasis of reduced hybrid fitness that is largely indepen-dent of the environment in which hybrids are reared, i.e.,intrinsic postzygotic isolation (Orr and Presgraves 2000;Coyne and Orr 2004). Research on the genetics of in-trinsic postzygotic isolation has led to four interrelatedgeneralizations.
First and foremost, there is now a broad consensusthat intrinsically low hybrid fitness (not attributable toecological interactions) is often the result of deleteriousepistatic interactions between genes from divergentparental genomes (Coyne and Orr 2004; Noor andFeder 2006). These ‘‘Dobzhansky–Muller incompatibil-ities’’ (DMIs) can arise when an ancestral population
(with genotype AABB) diverges into two isolated pop-ulations that fix different derived alleles (aaBB andAAbb, respectively). The derived alleles a and b must becompatible with the ancestral alleles A and B, or else thenew mutations would not have been able to pass througha heterozygous state to become fixed. However, there isno guarantee that the derived alleles are compatible witheach other, so there is a chance that in hybrids (AaBb)the derived alleles a and b interact (or fail to interact) ina way that reduces hybrid fitness (Dobzhansky 1934;Muller 1939; Orr 1993, 1995; Turelli and Orr 1995,2000; Orr and Turelli 2001). Several incompatibilityloci have been identified (Tinget al. 1998; Barbash et al.2003; Presgraves 2003; Presgraves et al. 2003), al-though very few pairwise interactions have been fullydescribed (Schartl et al. 1999; Brideau et al. 2006)
A second generalization, Haldane’s rule, states thatwhen one sex is inviable or infertile in hybrids, it is theheterogametic sex (e.g., XY or ZW) (Haldane 1922).This rule holds for the vast majority of taxa surveyed todate (Laurie 1997; Coyne and Orr 2004). It is widelyaccepted that this pattern arises for hybrid viabilitybecause DMIs tend to be recessive: incompatibilities
1Corresponding author: Section of Integrative Biology, One UniversityStation C0930, University of Texas, Austin, TX 78712.E-mail: [email protected]
Genetics 178: 1037–1048 (February 2008)
between autosomal alleles a and b will generally bemasked by the successful (dominant) interaction be-tween the compatible ancestral alleles A and B. How-ever, incompatibilities involving an X-linked locus willbe hemizygous in the XY sex and hence can reducefitness even if they are recessive (Muller 1942; Turelli
1998). These hemizygous incompatibilities also contrib-ute to a third generalization, the large-X effect (or‘‘Coyne’s rule’’), which observes that X chromosomesmake a disproportionate contribution to heterogameticinviability and/or sterility (Coyne and Orr 1989a, 2004;Turelli and Orr 2000).
The past few years have seen a revival of interest in afourth trend, ‘‘isolation asymmetry,’’ in which reciprocalhybrid crosses yield different degrees of hybrid inviabil-ity or sterility (Tiffin et al. 2001; Bolnick and Near
2005; Turelli and Moyle 2007). These asymmetrieshave been documented in plants, fungi, insects, andvertebrates (Sturtevant 1920; Muller 1942; Thornton
1955; Oliver 1978; Harrison 1983; Wu and Davis 1983;Rakocinski 1984; Coyne and Orr 1989b; Gallant andFairbain 1997; Presgraves and Orr 1998; Navajas et al.2000; Tiffin et al. 2001; Willett and Burton 2001;Presgraves 2002; Dettman et al. 2003). In 14 diversegenera of angiosperms, Tiffin et al. (2001) found asym-metric postmating isolation in 35–45% of species pairs. Ina freshwater fish family, Centrarchidae, 18 of 20 speciespairs with reciprocal cross viability data exhibited asym-metries (Bolnick and Near 2005). For instance, crossesbetween the congeners Lepomis gulosus and L. macrochirusyield average hybrid survival rates of 77% (L. gulosus dam)and 35% (L. macrochirus dam). This F1 hybrid asymmetryhas recently been dubbed ‘‘Darwin’s corollary to Hal-dane’s rule’’ by Turelliand Moyle (2007), in recognitionof Darwin’s own detailed attention to the pattern (Darwin
1859, Chap. 8).Asymmetric hybrid incompatibilities cannot be ex-
plained by traditional autosome–autosome DMIs. Thisis because F1 hybrid autosomal genotypes will be thesame regardless of cross direction (AaBb); note that thissymmetry does not hold for subsequent backcross prog-eny (Welch 2004) or in cases in which taxa differ inautosomal gene content (for simplicity, we ignore thiscomplication below). F1 asymmetries can be explainedby DMIs involving uniparentally inherited genetic fac-tors (Turelli and Moyle 2007). Four major classes ofasymmetrical DMIs have been discussed. First, Haldane’srule may be asymmetrical because one species’ X chro-mosome may have different numbers or magnitudes ofincompatibilities than the other species’ X. That is, hy-brids with genotype X1YAa need not have equal fitnessto X2YAa hybrids (alternatively, the sex chromosomescan interact asymmetrically). However, these asymmetrieswill be observed only in the heterogametic sex. Second,asymmetries may arise from DMIs between mitochond-rially encoded genes and nuclear genes (Kenyon andMoraes Carlos 1997; Rand et al. 2001, 2004; Rawson
and Burton 2002; Sackton et al. 2003; Burton et al.2006; Harrison and Burton 2006). Incompatibilitiesinvolving genes in cytoplasmic organelles may be asym-metric because, as with Haldane’s rule, one species’mitochondria (or chloroplasts) may contribute to moreor stronger DMIs (i.e., the fitness of M1Aa need notequal the fitness of M2Aa). We refer to this category ofDMI as a ‘‘mito-nuclear incompatibility,’’ to distinguishit from the third possible cause of asymmetry, arisingfrom maternal–zygotic incompatibilities. Maternal–zygoticincompatibilities can occur when maternally encodedtranscription factors in the oocyte fail to properly regu-late expression of paternally derived genes (Whitt et al.1977; Philipp et al. 1983; Turelli and Moyle 2007). Inthis case the DMI is not between genes within the hy-brid, but between the hybrid and maternal genomes. Afourth category of asymmetry arises from gametophyte–sporophyte interactions in plants (Tiffin et al. 2001;Turelli and Moyle 2007).
Turelliand Moyle (2007) modeled the evolution ofreciprocal hybrid asymmetry. In their analysis, they madea crucial distinction between stochastic and determin-istic contributions to asymmetry. To understand this dif-ference, consider mito-nuclear DMIs. First, let us assumethat both species are equally likely to substitute novelmitochondrial haplotypes and that the mitochondrialdifferences are equally likely to generate an incompat-ibility with the other species’ nuclear genome. Even so,the stochastic nature of substitutions and DMI occur-rence makes it very likely that one hybridization direc-tion accumulates more potent mito-nuclear DMIs thanthe other, yielding asymmetric hybrid viability. In con-trast to this purely stochastic explanation, it is possiblethat there is a systematic bias as to which species de-velops more mitochondria-based DMIs. This systematicbias arises when there is an asymmetry in the relativesubstitution rates of the two populations: whicheverspecies has a relatively faster rate of mitochondrial evo-lution will carry more substitutions and hence have agreater capacity for incompatibilities with the other spe-cies’ nuclear genome. By ‘‘relatively faster,’’ we mean thatthe mitochondrial genome experiences an acceleratedsubstitution rate relative to its nuclear genome substitu-tion rate (Turelli and Moyle 2007). If instead bothparts of the genome accelerate equally, one species’mitochondria has more opportunities to pick up in-compatibilities with the other species’ nuclear genome,but the reciprocal is also true so no systematic asymme-try ensues.
Turelli and Moyle’s (2007) analyses led them toconclude that ‘‘unless relative rate differences are ex-treme . . . , stochastic effects are more likely to explain ob-served levels of postmating asymmetry than systematicinterspecific differences in the relative rates of molec-ular evolution’’ (Turelli and Moyle 2007, p. 1068).Specifically, they found that levels of asymmetry (i.e., themagnitudes of reciprocal cross differences) are likely to
1038 D. I. Bolnick et al.
have an appreciable deterministic basis only if one spe-cies experiences at least a fourfold acceleration in itsmitochondrial evolution relative to the other species,while the nuclear genomes remain at their previousrate. However, Turelli and Moyle (2007, p. 1081) didpropose that ‘‘if systematic rate differences . . . do play asignificant role in postmating asymmetry, then thedirection of asymmetry should be predictable fromthe relative rates of evolution for the loci that contributeto ½unidirectional� DMIs.’’ Here, we extend the analysisof Turelli and Moyle, focusing on the direction of asym-metries (i.e., which maternal parent is more likely toyield lower hybrid viability). Our analysis shows thatmoderate differences in rates of mitochondrial evolu-tion can have an appreciable impact on which species islikely to be the ‘‘worse’’ maternal parent in reciprocalcrosses. We then test this prediction by comparing thedirections of F1 viability asymmetry, and mitochondrialand nuclear substitution rates, in a clade of NorthAmerican freshwater fishes (Centrarchidae). We find aweak but statistically significant tendency for the specieswith faster mitochondrial evolution to be the worsematernal parent in reciprocal hybrid crosses. Together,our theory and data suggest that systematic effectsinvolving mito-nuclear interactions may frequently con-tribute to ‘‘Darwin’s corollary.’’
MODEL
Given that our data on hybrid asymmetry in cen-trarchids concern embryo-to-larval viability and relativerates of mitochondrial vs. nuclear evolution, we focuson analyzing hybrid inviability and mito-nuclear DMIs.However, we emphasize that many asymmetrical DMIsmay involve maternal effects (cf. De Renzis et al. 2007),and there is some evidence for maternal effects con-tributing to hybrid asymmetry in centrarchids (Whitt
et al. 1977; Philipp et al. 1983). Following Turelli andOrr (1995, 2000), the asymmetry analysis of Turelli
and Moyle (2007) assumes that individual DMIs con-tribute additively to a ‘‘hybrid breakdown score,’’ de-noted S, and that hybrid viability decreases as S increasesup to a threshold value, denoted C, beyond which hy-brids become completely inviable. Let Sij denote thebreakdown score for hybrids between species i and j thathave species i mothers.
Considering only mito-nuclear DMIs, the existence ofa preferred direction for asymmetry is determined bythe parameter
d1 ¼ y1 � y1; ð1Þ
where y1 (y1) is the fraction of the expected mitochon-drial (nuclear) substitutions that occur in lineage 1. Theparameter d1 thus represents the degree of asymmetryin rates of mitochondrial evolution, relative to nuclearrates. This quantity can be approximated from estimates
of the number of mtDNA and nuclear substitutionsoccurring in each lineage, using a rooted phylogeny. Forinstance, in Figure 1 more than half of the mitochon-drial substitutions have occurred in the lineage leadingto species 1, so y1¼B(m)1/½B(m)1 1 B(m)2�. 0.5, whereB(m)i is the mitochondrial branch length from species ito the most recent common ancestor of a given speciespair. Turelliand Moyle (2007) showed that when d1 .
0, corresponding to a higher proportion of mitochon-drial vs. nuclear substitutions occurring in lineage 1, theexpected hybrid breakdown score is higher for crosseswith species 1 as the maternal parent ½E(S12) . E(S21)�.This result arises because crosses with taxon 1 motherswill on average experience more mito-nuclear DMIsthan do the reciprocal crosses. Rather surprisingly, theyfound that the level of reciprocal-cross asymmetry, asquantified by the absolute value of the hybrid viabilitydifference between reciprocal crosses, is not appreciablyaffected by d1 . 0, unless the relative rate differences areextreme, for instance, d1 . 0.3. If we assume that bothlineages experience equal rates of nuclear substitutions,d1 ¼ 0.3 corresponds to lineage 1 experiencing on av-erage four times as many mitochondrial substitutions aslineage 2. However, Turelli and Moyle did not investigate
Figure 1.—An illustration of the deterministic hypothesisto explain asymmetric F1 viability. For any two species withina larger phylogeny, one can estimate the branch length fromeach species back to its common ancestor (arrow indicatesoutgroup rooting) for both mitochondrial loci (solid lines,m) and nuclear loci (dotted lines, n). In this schematic, spe-cies 1 has accelerated mtDNA evolution (relative to nuclearrates), resulting in d1 . 0. Theory predicts that species 1 thushas more potential for creating genetic incompatibilities inhybrids when it is the maternal parent.
Asymmetric Postmating Isolation in Centrarchid Fishes 1039
how the probability of directional asymmetry, e.g., P(S12 .
S21), depends on d1.As shown below, the probabilities of directional
asymmetry, like the absolute magnitudes of asymmetry,depend on divergence time, t. We measure divergencetimes in scaled units, t ¼ t/TC, with TC defined as theharmonic mean of the times until the expected break-down scores from the reciprocal crosses reach thethreshold value C. In principle, three quantities mightbe used to quantify the probability of directionalasymmetry:
PðS12 . S21Þ; ð2Þ
PðS12 . S21; S21 , CÞ; ð3Þ
or
PðS12 . S21 j Smin , CÞ ¼ PðS12 . S21; S21 , CÞPðSmin , CÞ : ð4Þ
The first, (2), is simply the probability that the breakdownscore for hybrids with taxon 1 mothers is larger than thatfor hybrids with taxon 2 mothers. Because the expectednumber of DMIs is proportional to t2 (Orr 1995; Orr andTurelli 2001), differences between the expected valuesof S12 and S21 will become increasingly large as divergencetime (t) increases; and P(S12 . S21) will approach 1 whend1 . 0 and t is large. However, this will not be manifest infitness differences once t is sufficiently large that bothS12 and S21 exceed C, and all hybrids become inviable. Thissuggests measure (3), which requires that S12 . S21 butS21 , C, so that a fitness difference is observedbetween thehybrids from reciprocal crosses. However, this quantitymust approach 0 as t increases, because the conditionS21 , C will be violated eventually. Hence, we suggest thatthe time dependence of the probability of directionalasymmetry may be most informatively quantified by (4),which conditions on observed fitness asymmetry andreports the probability that the observed asymmetry goesin a specified direction (e.g., lower viability with taxon 1mothers).
In F1 hybrids, there is an important difference be-tween mito-nuclear DMIs and DMIs between autosomalloci. Whereas both autosomal loci involved in a DMIare heterozygous for incompatible alleles, mito-nuclearDMIs involve hemizygous loci interacting with hetero-zygous loci and are expected to have a systematicallylarger effect. Indeed this distinction underlies Muller’s(1942) ‘‘dominance theory’’ for Haldane’s rule. Theexpected ratio of effects of autosomal vs. mito-nuclearDMIs is denoted h0 and, as discussed by Turelli andOrr (2000) and Turelli and Moyle (2007), is ex-pected to be small, on the order of 0.1. As shown byTurelli and Moyle (2007), the quantitative propertiesof reciprocal-cross asymmetry can be adequately described
using a bivariate normal approximation for the recipro-cal breakdown scores (S12, S21). For realistic parametervalues, the behavior of this distribution depends on h0
only through a composite parameter, denoted h, thatquantifies the relative effect of nuclear (symmetrical) vs.mito-nuclear (asymmetrical) DMIs. With h¼ 0, all DMIsare mito-nuclear and the level of expected asymmetry isgreatest. Whereas with h¼ 10, symmetrical nuclear DMIscontribute on average 10 times as much to the inviabilityof hybrids as mito-nuclear DMIs, and little asymmetry isexpected. Under the bivariate Gaussian approximation,the probabilities in (4) would be calculated by numer-ical integration using Mathematica 6.0 (Wolfram 2003).
Here we illustrate how these probabilities of direc-tional asymmetry (Equation 4) vary with: divergencetime d1; C, the average number of mito-nuclear DMIsneeded to produce complete postmating isolation; CV,the coefficient of variation of DMI effects; and h, whichmeasures the relative effect of symmetrical vs. asymmet-rical DMIs. To explain the levels of asymmetry observedin centrarchid fishes by Bolnick and Near (2005),Turelli and Moyle (2007) argued that C and h mustboth be fairly small, with C on the order of 5 and h onthe order of 1. We focus on these values. For additionaldiscussion of the model parameters and their biologicalinterpretation, see Turelli and Moyle (2007).
Figure 2, A–D, shows how the probability of aspecified direction of asymmetry, PðS12 . S21 j Smin , CÞ,varies as a function of divergence time (t ¼ t/TC) whilevarying the values of d1 (A), C (B), CV (C), and h (D),around base values of d1¼ 0.2, C¼ 5, CV¼ 0.5, and h¼1, while holding h0 ¼ 0.1. Figure 2A shows, as expected,that the probability of directional asymmetry increaseswith the asymmetry of mitochondrial substitution rates,d1. The case d1 ¼ 0 provides a control, showing that theprobability of a given direction of asymmetry is 0.5 forall divergence times, as expected. When mitochondrialrates are equal, either species is equally likely to be thematernal parent for the less viable cross direction. As d1
increases, species 1 exhibits faster mitochondrial evolu-tion and hence is increasingly likely to produce lessviable hybrids when acting as the maternal parent. Theresults with d1 ¼ 0.3 are notable. This corresponds tomtDNA evolution being four times faster in lineage 1than in lineage 2, but it produces a very small effect onthe expected magnitude of asymmetry (see Figure 6C ofTurelli and Moyle 2007). In contrast, there is anappreciable effect on the probability of directionalasymmetry, which rises from�0.65 to 0.77 as t increasesfrom 0.6 to 1.2. Over this range of divergence times, theprobability that at least one of the reciprocal crossesproduces viable hybrids ½i.e., the denominator of (4)�decreases from very near 1 to 0.33.
Figure 2B varies C, the average number of mito-nuclear DMIs needed to produce complete inviability,holding d1 ¼ 0.2, h ¼ 1, h0 ¼ 0.1, and CV ¼ 0.5. Formoderate divergence times, corresponding to signifi-
1040 D. I. Bolnick et al.
cant probabilities of hybrid inviability, the probability ofdirectional asymmetry approaches 1 as C becomes large.This reflects the fact that as C increases, the stochasticprocess of DMI accumulation becomes increasinglydeterministic. (As shown by Orr and Turelli 2001,the variance in the number of DMIs is close to the mean;so the coefficient of variation of breakdown scoresproducing high levels of inviability decreases as Cincreases.) As shown by Turelli and Moyle (2007),when C is $10, we expect much lower magnitudes ofasymmetry than observed in the centrarchid data ofBolnick and Near (2005).
Figure 2C varies CV, the coefficient of variation ofDMI effects, with d1 ¼ 0.2, C ¼ 5, h0 ¼ 0.1, and h ¼ 1. Itshows that for realistic values of CV (namely, CV , 1),the magnitude of CV has a negligible effect on theprobability of directional asymmetry. As expected, as CVincreases, the probability of directional asymmetrydecreases slightly as this additional source of stochas-ticity masks the deterministic signal associated with d1 .
0. Figure 2D varies h, the ratio of the expected con-tribution of symmetrical vs. asymmetrical DMIs tohybrid inviability, holding C ¼ 5, d1 ¼ 0.2, h0 ¼ 0.1,and CV ¼ 0.5. With very large h, almost all DMIs are
symmetrical and reciprocal-cross asymmetry should benegligible, in contrast to the considerable asymmetrydocumented by Bolnick and Near (2005). Figure 2Dshows that the probability of directional asymmetry, likeits expected magnitude, increases with decreasing h.
In conclusion, while asymmetric rates of mitochon-drial evolution (d1 . 0) have a negligible effect on themagnitude of asymmetries in reciprocal F1 crosses(Turelli and Moyle 2007), the direction of asymmetryis sensitive to d1. We therefore predict that for a givenpair of species, the maternal parent species with rela-tively faster mitochondrial evolution will tend to pro-duce less viable offspring. Below we present a test of thisprediction using data from centrarchid fishes. However,it is necessary to clarify the connection between our pa-rameters d1 and h and our mtDNA data. As noted in theIntroduction, mito-nuclear DMIs are not the only po-tential source of postmating reproductive asymmetry. Inparticular, the data of Whitt et al. (1977) and Philipp
et al. (1983) indicate that maternal–zygotic interactionsare also likely to be important. Hence, mito-nuclearinteractions probably make up only a portion of theasymmetrical DMIs that contribute to h and the relevantvalue of d1 is likely to differ from estimates based on
Figure 2.—The probability that a specified reciprocal-cross direction yields lower hybrid fitness, given that reciprocal-crossasymmetry is observed; i.e., PðS12 . S21 j Smin , CÞ, as a function of scaled divergence time t/TC. (A) The effect of varying the asym-metry in relative rates of mitochondrial evolution, d1: d1 ¼ 0 (dashed line), 0.05 (gray–blue), 0.1 (blue), 0.2 (green), 0.3 (lightblue), and 0.4 (red), with C ¼ 5, h ¼ 1, h0 ¼ 0.1, and CV ¼ 0.5. (B) The effect of varying the average number of mito-nuclearincompatibilities C at which inviability is complete: C ¼ 5 (gray–blue), 10 (blue), 20 (green), and 100 (red), with d1 ¼ 0.2, h ¼ 1,h0¼ 0.1, and CV¼ 0.5. (C) The effect of varying the coefficient of variation of DMI effects, CV: CV¼ 0 (gray–blue), 0.5 (blue), and1.0 (red), with C ¼ 5, d1 ¼ 0.2, h ¼ 1, and h0 ¼ 0.1. (D) The effect of varying the relative contribution of symmetrical vs. asym-metrical DMIs h: h ¼ 0.1 (red), 1.0 (gray–blue), and 5.0 (green), with C ¼ 5, d1 ¼ 0.2, h0 ¼ 0.1, and CV ¼ 0.5.
Asymmetric Postmating Isolation in Centrarchid Fishes 1041
mtDNA alone. To the extent that the other uniparentallyinherited factors evolve independently of mtDNA (i.e.,do not tend to accelerate with mtDNA acceleration),our mtDNA estimates are likely to correctly estimate thesign of d1 but overestimate its magnitude. We simply donot have enough data to predict how far above 0.5 theprobabilities of directional asymmetry are likely to be.However, our best guesses for the parameters suggest thatvalues .0.6 are unlikely.
METHODS
Hybrid viability data and calculating reciprocalasymmetry: We compiled published data on hybridviability among centrarchid species (see Bolnick andNear 2005 for details). Centrarchids are externalfertilizers, so it is possible to strip eggs and sperm frommature individuals to generate in vitro hybrids. Viability(Vij) of a particular cross direction (species i and j asdam and sire, respectively) was measured as the per-centage of hybrid embryos that hatched, divided by thehatch rate of intraspecific embryos from the same clutchof eggs, to control for egg maturity. Note that Vij canexceed 100% if hybrids exhibit higher hatching ratesthan embryos from intraspecific crosses. Fertilizationrates are consistently high (.90%) even for distantlyrelated species pairs and are not correlated with evolu-tionary divergence (West and Hester 1966; Merriner
1971), so embryo hatch rates are a good measure ofviability.
The compiled data set contains 18 species pairs withreciprocal-cross data, not including 3 reciprocal pairswith zero viability in both directions (Table 1). For eachof the 18 pairs, we identified which species was the ‘‘worsedam,’’ i.e., the species i for which Vij , Vji. Note thatmany of the published reciprocal crosses do not reportsufficient information to determine whether reciprocal-cross viabilities are significantly different. However, largeclutch sizes (often ?500) mean that even slight differ-ences in viability are usually statistically significant. Allnine reciprocal crosses for which sample sizes were avail-able exhibited significant asymmetries (see Bolnick andNear 2005 for details). Note also that centrarchids donot appear to follow Haldane’s rule (Bolnick and Near
2005), consistent with the finding that they have nokaryotypically distinct sex chromosomes (Roberts 1964).In an intraspecific cross of L. cyanellus, only 1 of 429polymorphic amplified fragment length polymorphism(AFLP) markers was fully linked to sex (present in allfemales and in no males; Lopez-Fernandez and Bolnick
2007). Consequently, asymmetric hybrid viabilities areunlikely to have arisen via Haldane’s rule and are morelikely to reflect mito-nuclear or maternal–zygotic incom-patibilities (Whitt et al. 1977; Philipp et al. 1983).
Phylogenetic analyses and branch length estimation:We sequenced five loci, totaling 5533 bp, from at leastone individual of all 32 described centrarchid species
(Near et al. 2004, 2005). The loci include multiplemitochondrial genes (ND2, 16S, and a set of tRNAs) andfour nuclear genes (rhodopsin, TMO4C4, and intronsfrom calmodulin and S7 ribosomal protein). We obtaineda consensus phylogeny using a partitioned Bayesian anal-ysis (see Near et al. 2005 for details). Similar phylogenieswere found using maximum-likelihood and parsimonymethods (Near et al. 2004, 2005). Phylogenies estimatedfrom individual loci were often not as well resolved asthose resulting from analysis of the combined data set;however, there was little phylogenetic incongruenceamong the trees estimated from each locus. The oneexception concerned Micropterus treculi, which mtDNAand nuclear gene phylogenies place in very differentlocations within the Micropterus clade. We omitted thisspecies from the present molecular analyses, but itsinclusion has no effect on our results.
We obtained branch lengths for each locus by estimat-ing their substitution rates on a constrained topology,the Bayesian consensus tree (Figure 3). The optimal like-lihood model was obtained for each nuclear locus andmitochondrial gene using ModelTest 3.0 (Posada andCrandall 1998), and branch lengths (the expectednumber of substitutions per nucleotide site) were esti-mated using these specific maximum-likelihood modelsin PAUP* 4.0b10 (Swofford 2002). For each speciespair, these branch lengths represent Bij, the number ofsubstitutions from each species i to the most recentcommon ancestor of the pair, for locus j (Figure 1).Whichever species has a larger Bij has a faster rate ofmolecular evolution at that locus. We summed thebranch lengths across all mitochondrial loci to obtain atotal mitochondrial branch length B(m)i for each mem-ber of each species pair. Similarly, we summed acrossnuclear loci to obtain total nuclear branch lengthsB(n)i. Next, we calculated the rate asymmetry formitochondrial (yi ¼ BðmÞi=
P2i¼1 BðmÞi) and nuclear
(yi ¼ BðnÞi=P2
i¼1 BðnÞi) loci. Finally, we calculated therelative mitochondrial rate asymmetry di (Equation 1,Figure 1) for each species in each of the 18 species pairs.In a given species pair, the species with the positive di isinferred to have accelerated mitochondrial evolution,and the other species of the pair will have a negative di ofequal magnitude. Note that the theoretical predictionsconcern overall substitution rates, rather than rates ofnonsynonymous substitutions per se. Furthermore, we donot necessarily expect that the particular genes usedhere are the causal agents in asymmetric incompatibil-ities. We therefore focus on overall substitution rates,rather than nonsynonymous substitutions.
To assess the robustness of our results, we used a sec-ond approach to calculating di. We calculated yi andyi separately for each locus, then averaged across loci toobtain a mean mitochondrial and mean nuclear yi andyi , and took the difference to obtain di. Our results werequalitatively identical with this method, so we focus on thefirst formulation, which more closely approximates the
1042 D. I. Bolnick et al.
TA
BL
E1
Dat
ao
nre
cip
roca
lcr
oss
esam
on
gce
ntr
arch
idsp
ecie
s
Spec
ies
pai
rH
ybri
dvi
abil
ity
(%o
fco
ntr
ol
cro
ss)
Rel
ativ
evi
abil
ity
dif
fere
nce
Wo
rse
dam
Rat
eas
ymm
etry
di
Spec
ies
wit
hd
i.
0
Wo
rse
dam
:fa
ster
mtD
NA
?
Smal
ler
spec
ies
No
de
AB
A3
BB
3A
AB
Spec
ies
Fas
ter
mtD
NA
?W
ors
ed
am?
AA
.ru
pest
ris
L.
gulo
sus
25.
5�
0.63
6A
0.00
2�
0.00
2A
Yes
BN
oN
oA
A.
rupe
stri
sL
.m
acro
chir
us
265
0.80
8B
�0.
294
0.29
4B
Yes
BYe
sYe
sB
A.
rupe
stri
sP.
nig
rom
acu
latu
s41
86.5
�0.
526
A0.
029
�0.
029
AYe
sA
Yes
Yes
IL
.cy
anel
lus
L.
gulo
sus
9910
7�
0.07
5A
0.01
2�
0.01
2A
Yes
AYe
sYe
sH
L.
cyan
ellu
sL
.m
acro
chir
us
105
900.
143
B0.
045
�0.
045
AN
oE
qu
alN
AN
AG
L.
cyan
ellu
sL
.m
icro
loph
us
118
107
0.09
3B
�0.
104
0.10
4B
Yes
BYe
sYe
sD
L.
cyan
ellu
sM
.sa
lmoi
des
1180
�0.
863
A0.
062
�0.
062
AYe
sA
Yes
Yes
HL
.gu
losu
sL
.m
acro
chir
us
7735
0.54
5B
0.04
5�
0.04
5A
No
AYe
sN
oG
L.
gulo
sus
L.
mic
rolo
phu
s10
782
0.23
4B
0.22
4�
0.22
4A
No
BN
oYe
sD
L.
gulo
sus
M.
salm
oide
s75
103
�0.
272
A0.
062
�0.
062
AYe
sA
Yes
Yes
AL
.gu
losu
sP.
nig
rom
acu
latu
s1.
72.
3�
0.26
1A
�0.
155
0.15
5B
No
AN
oYe
sG
L.
mac
roch
iru
sL
.m
icro
loph
us
7710
2�
0.24
5A
�0.
144
0.14
4B
No
BYe
sN
oD
L.
mac
roch
iru
sM
.sa
lmoi
des
770
�0.
900
A0.
324
�0.
324
AYe
sA
Yes
Yes
AL
.m
acro
chir
us
P.n
igro
mac
ula
tus
1344
�0.
705
A0.
290
�0.
290
AYe
sA
Yes
Yes
DL
.m
icro
loph
us
M.
salm
oide
s0
43�
1.0
A0.
399
�0.
399
AYe
sA
Yes
Yes
EM
.do
lom
ieu
M.
salm
oide
s82
102
�0.
196
A0.
058
�0.
058
AYe
sA
Yes
Yes
FM
.fl
orid
anu
sM
.sa
lmoi
des
109
910.
165
B�
0.12
10.
121
BYe
sA
No
No
CP.
ann
ula
ris
P.n
igro
mac
ula
tus
9877
0.21
4B
�0.
003
0.00
3B
Yes
BYe
sYe
s
No
de
iden
titi
esco
rres
po
nd
toF
igu
re3.
Fo
rea
chsp
ecie
sp
air
we
arb
itra
rily
assi
gna
spec
ies
Aan
dB
,wh
ich
we
use
tosp
ecif
yth
evi
abil
itie
so
fth
ere
spec
tive
reci
pro
calc
ross
es(f
emal
ep
aren
tli
sted
firs
t).
Via
bil
itie
sar
eca
lcu
late
db
yd
ivid
ing
the
per
cen
tage
of
hyb
rid
emb
ryo
sth
atsu
rviv
eto
larv
alst
age
by
the
per
cen
tage
of
intr
asp
ecifi
cco
ntr
ol
cro
ssem
bry
os
that
surv
ive.
Rel
ativ
evi
abil
ity
dif
fere
nce
isca
lcu
late
das
the
dif
fere
nce
invi
abil
itie
s,st
and
ard
ized
by
the
hig
her
valu
e.T
he
‘‘wo
rse
dam
’’is
the
spec
ies
pro
vid
ing
the
eggs
for
the
low
ervi
abil
ity
reci
pro
cal
cro
ss.
Rel
ativ
em
ito
cho
nd
rial
rate
s,d
i,fo
rb
oth
spec
ies
are
calc
ula
ted
fro
mb
ran
chle
ngt
hd
ata
insu
pp
lem
enta
lT
able
1o
fth
esu
pp
lem
enta
lA
pp
end
ix(h
ttp
://w
ww
.gen
etic
s.o
rg/
sup
ple
men
tal/
).T
he
spec
ies
wit
hth
eh
igh
erva
lue
exh
ibit
sre
lati
vely
fast
erm
tDN
Aev
olu
tio
n.
Un
der
the
hea
din
g‘‘S
mal
ler
spec
ies’
’w
eid
enti
fyth
esp
ecie
sw
ith
the
smal
ler
max
imu
mb
od
ysi
zean
dst
ate
wh
eth
erth
isco
inci
des
wit
hh
avin
gfa
ster
-evo
lvin
gm
tDN
Ao
rb
ein
gth
ew
ors
ed
am.
Asymmetric Postmating Isolation in Centrarchid Fishes 1043
theoretical values. Using either approach, we acknowledgethat this estimate is based on a very limited sample ofgenes. This limitation should introduce additional errorinto our analysis and reduce the likelihood of detectingan effect if one existed, so our empirical tests should tendto be conservative.
Testing for systematic causes of asymmetricreciprocal F1 inviability: On the basis of our model,we predicted that systematic asymmetries might arise be-cause, for a given species pair, the maternal parent withrelatively faster mitochondrial evolution (positive di)should produce less viable hybrids (Figure 1). We testedthis directional prediction with a one-tailed sign test. Wecounted the number of cases (X) out of N reciprocalcrosses where the species with di . 0 was also the worsedam (Vij , Vji), converted this into a fraction (X/N), andtested the null hypothesis that X/N ¼ 0.5 using abinomial (sign) test. However, the 18 species pairs arenot phylogenetically independent. We therefore ap-plied a version of Coyne and Orr’s (1989b) node-basedphylogenetic correction. Each species pair was assigneda score of 1 if the species with a positive di also was theworse dam and 0 if the theoretical expectation was notmet. We then averaged the scores across all species pairsthat diverged at particular node, leaving us with ninenode-averaged scores. This averaging used the node-weighted approach of Fitzpatrick (2002), althoughresults were equivalent when we used the nonweightedapproach of Coyne and Orr (1989b). A t-test thenevaluated the null hypothesis that scores averaged 0.5,
against a directional alternative hypothesis that scorestend to exceed 0.5.
Another way to test for the same trend is to regressviability asymmetry against mitochondrial rate asymme-try. If the species with faster-evolving mitochondria isthe worse maternal parent, we also expect that the rateasymmetry measure is negatively correlated with relativeviability. We measured relative viability for species 1 as
RV1 ¼V12 � V21
MaxðV12;V21Þ; ð5Þ
where Vij is the viability of hybrids with species i and j asthe maternal and paternal parents, respectively. Arbi-trarily focusing on one species in each pair, di (fastermitochondrial evolution) should be negatively corre-lated with relative viability RVi. Regressing RVi against di
is redundant with the sign test (above), but illustratesthe robustness of our results to another statistical method.To apply phylogenetic corrections, we calculated the meandi and RVi for each node and repeated the regression. Incalculating mean di and RVi, we randomly selected oneof the sister clades to be the source of the focal speciesfor each species pair.
While we expected that mitochondrial rate asymme-try affects the direction of viability asymmetry, Turelli
and Moyle (2007) predicted that the magnitude ofviability asymmetry should be relatively insensitive to di.To test the latter relationship, we calculated the absolutemagnitude of rate and viability asymmetries for each
Figure 3.—A phyloge-netic hypothesis for theCentrarchidae, based onDNA sequences from threemtDNA gene regions andfour nuclear genes. Bayes-ian posterior probabilitiesare provided for all nodes.See Near et al. (2005) fordetails.
1044 D. I. Bolnick et al.
species pair. The magnitude of asymmetry was calcu-lated by taking the absolute value of di and RVi. Weregressed jRVij against jdij, using nodes as the level ofreplication (averaging multiple species pairs at a givennode). No significant relationship is expected.
Comparative studies such as this one are fundamen-tally limited in their ability to infer causation, relying oncorrelations across species. When correlations betweentwo variables do occur, they might be explained by mu-tual correlation with a third variable. Previous analysessuggested a possible association between asymmetricviability and adult body size, in which the smaller speciestends to be the worse dam (Bolnick et al. 2006). To eval-uate whether body size might represent a confoundingvariable, we conducted two phylogenetically correctedsign tests (t-tests on averaged 0/1 scores for each node),to test (1) whether the smaller species is the worse damand (2) whether the smaller species tends to have fastermitochondrial evolution.
RESULTS
A sign test confirmed that species with faster-evolvingmitochondria tend to be the worse dam, producing F1
hybrids with lower viability than the reciprocal crosswith a slow-mitochondria female parent (Table 1). Of 18species pairs, 13 pairs matched our expectation, with aone-tailed binomial probability P ¼ 0.0481 under thenull hypothesis of no association between mtDNA rateand asymmetry. A one-tailed test is appropriate because
we are evaluating an a priori directional hypothesis. Thephylogenetically corrected t-test also confirmed that nodeaverages of the sign test scores consistently exceeded thenull expectation of 0.5 (mean node averages¼ 0.79, t9¼2.32, P ¼ 0.024). Both the sign test and the phylogenet-ically corrected t-test are consistent with our modelshowing that that variation in mitochondrial evolutionrates generates systematic differences in reciprocal F1
hybrid viabilities. The theory also emphasized that asym-metric viability should result from asymmetric mitochon-drial rates relative to nuclear rates (d), not mitochondrialasymmetry alone (n) (Turelli and Moyle 2007). Con-sistent with this expectation, we found no associationbetween the direction of viability asymmetry and n: thespecies with the larger n was also the worse dam in only7 of 18 cases.
A quantitative comparison of viability and molecularevolution rate asymmetries is also consistent with adeterministic contribution to Darwin’s corollary. Therewas a significant negative correlation between the relativeviability difference and the difference in relative mtDNArates (F1,16¼ 10.11, P¼ 0.006, r2¼ 0.387). The regressionwith node averages also supported a negative correlation(Figure 4, F1,7¼ 8.392, P¼ 0.023, r2¼ 0.545). This trendconfirms the sign test results because it incorporates thedirection of the asymmetries. That is, when species A hasthe faster evolution rate (positive rate difference), it isalso the worse dam (negative relative viability). In con-trast to results for the direction of asymmetry, we foundno significant correlation between mitochondrial rateasymmetry and the magnitude of viability asymmetry (Fig-ure 5, regression of node averages: F1,7¼ 3.45, P¼ 0.106,r2 ¼ 0.33), although the trend was positive.
The association between mitochondrial rates and via-bility asymmetry is admittedly a correlation, rather thanevidence of causation. Highlighting this fact, we also
Figure 4.—The quantitative relationship between the mito-chondrial substitution rate of asymmetry d1 and the degree ofasymmetry in F1 viability, using node averages. Both relativerate and relative asymmetry are calculated, selecting species1 for each pair from an arbitrary clade subtending each node,allowing both positive and negative values of each variable.This regression is partially redundant with the sign test de-scribed in the text, as it tests for effect direction: this is illus-trated by dividing the figure into quadrants (dotted lines).The one node that runs counter to our expectation is indi-cated by an open rather than a solid circle.
Figure 5.—The quantitative relationship between the mag-nitude (rather than direction) of mitochondrial rate asymme-try d1 and the magnitude of viability asymmetry, using nodeaverages.
Asymmetric Postmating Isolation in Centrarchid Fishes 1045
found that the worse maternal parent tended to havethe smaller body size in a given pair (sign test on raw data,13/17 cases, P ¼ 0.049; t-test on node averages, meanscore¼ 0.796, t7¼ 2.53, P¼ 0.017; note that one speciespair was omitted due to identical maximum body sizes).The smaller species also has a marginally nonsignificanttendency to exhibit accelerated mitochondrial evolu-tion (sign test on raw data, 13/17 cases, P ¼ 0.049; t-teston node averages, mean score ¼ 0.712, t7 ¼ 1.51, P ¼0.085).
DISCUSSION
Asymmetric viability of reciprocal hybrid crosses islikely to arise from DMIs between uniparentally in-herited factors and autosomal biparentally inheritedloci (Turelli and Moyle 2007). Uniparentally inher-ited factors include hemizygous sex chromosomes, mito-chondria, and cytoplasm. These confer asymmetriesbecause when one species’ mitochondria have a delete-rious interaction with an autosomal locus from the otherspecies, there is no reason why the second species’ mi-tochondria must also confer incompatibility, or if it doesthere is no reason why their effects need be the same.Consequently, Coyne and Orr (2004, p. 310) stated that‘‘asymmetry provides prima facie evidence for the roleof the cytoplasm in postzygotic isolation.’’
Substitutions generating these uniparental DMIs(UDMIs) are expected to occur stochastically in bothparental species. Consequently, the direction of theasymmetry (which reciprocal cross is less viable) may beentirely random. However, there may also be a de-terministic contribution if one species is more likely toacquire a UDMI, for instance, if it exhibits acceleratedmitochondrial evolution and so is likely to harbor moresubstitutions. Our theoretical results predict that thedirection of F1 viability asymmetries depends on relativerates of mitochondrial evolution in the parental taxa.Our empirical results are consistent with this prediction:hybrid viability tends to be lower when the species withaccelerated mtDNA evolution is the maternal parent.This trend was confirmed by both a sign test and linearregression, with both raw and phylogenetically correcteddata. However, we find no evidence that the magnitudeof asymmetry is related to the difference in evolutionaryrates, consistent with previous theoretical predictions(Turelliand Moyle 2007). The agreement between ourtheory and empirical results suggests that mito-nuclearincompatibilities contribute to postzygotic reproductiveisolation in centrarchids, as has been previously demon-strated in other taxa ½e.g., Drosophila (Rand et al. 2001;Sackton et al. 2003) and intertidal copepods (Burton
et al. 2006)�. In particular, mito-nuclear incompatibili-ties may help explain the widespread asymmetricalviabilities of reciprocal crosses.
We emphasize two caveats regarding our results. First,the node-averaging approach to reducing phylogenetic
nonindependence does not guarantee full independenceof the data. For example, node averages may not be in-dependent if the same substitution(s) are involved inhybrid inviability between L. microlophus and L. cyanellus(Figure 3, node G) and between L. microlophus and M.salmoides (Figure 3, node D). At present there is no evi-dence that the same substitutions are involved in in-compatibility across different nodes, and the epistaticnature of DMIs makes this appear unlikely. Other moreformal approaches to phylogenetic correction are notapplicable here because they require branch lengths(e.g., Bolnick and Near 2005), which are an explana-tory variable in our study, or they are not equipped tohandle characters that are properties of pairs of taxarather than individual taxa ½e.g., independent contrasts(Felsenstein 1985)�.
A second caveat is that our results are correlative, as inany comparative analysis. We are not able to conclusivelyestablish causation, because we are not able to rule outthe possibility that the association we have found arisesfrom mutual dependence on an unidentified third vari-able. For instance, it is conceivable that both asymmetricviability and asymmetric mitochondrial branch lengthsare a result of differences in body size. We previouslynoted that for a given species pair, the smaller speciestends to be the worse maternal parent (Bolnick andNear 2005; Bolnick et al. 2006). In the present study, weconfirmed this trend with a phylogenetically correctedanalysis and also found that small body size exhibits amarginally significant tendency to be associated withaccelerated mitochondrial evolution. We are not, atpresent, able to rigorously tease apart the directions ofcausation in this triangle of correlations. However, weemphasize that theory provided an a priori predictionthat mitochondrial rate asymmetries should give rise toviability asymmetry, whereas there was no prediction re-garding body-size effects.
One possible explanation for the body-size effect isthat asymmetric hybrid viability could reflect a combi-nation of mito-nuclear and maternal–zygotic incompat-ibilities ½or genomic imprinting (Haig 2004)�. Althoughour data are consistent with a systematic effect drivingmito-nuclear UDMIs, this does not preclude other formsof incompatibilities acting simultaneously. Allozyme ex-pression studies in centrarchids have previously implicatedcyto-nuclear incompatibilities as a cause for asymmetry(Whitt et al. 1977; Philipp et al. 1983). When two spe-cies were crossed experimentally, F1 hybrids exhibiteddelayed onset of expression for a number of allozymes.When the maternally and paternally derived allozymeswere electrophoretically distinguishable, it was frequentlyfound that the maternal alleles were expressed normallybut the paternal alleles were delayed, even in reciprocalcrosses (Whitt et al. 1977). The simplest explanation isthat maternally encoded transcription factors in theoocyte cytoplasm were successfully regulating embry-onic expression of maternal alleles, but failing to regu-
1046 D. I. Bolnick et al.
late paternal alleles. The magnitude of delayed onset wascorrelated with the degree of hybrid inviability (Philipp
et al. 1983). Such cytoplasmic–nuclear incompatibilitiescould, in principle, be exacerbated by body size differ-ences between species. However, the most obvious rea-son for such an effect does not appear to hold: body sizeis not correlated with egg size (Bolnick et al. 2006), andegg size differences are not correlated with hybrid via-bility (Merriner 1971). However, evolutionary shifts indevelopment rate and other life-history characteristicsmay be associated with evolutionary shifts in either ma-ternally encoded cytoplasmic transcription factors or meta-bolic function, which could result in UDMIs.
The dominance theory for Haldane’s rule also relieson UDMIs, because it arises from incompatibilities be-tween a hemizygous and therefore uniparentally in-herited factor (e.g., the X chromosome in males) andautosomal loci. Asymmetric instances of Haldane’s ruleare widespread (Laurie 1997; Presgraves and Orr
1998), but these asymmetries are restricted to the hetero-zygous sex. There are several reasons why we believeHaldane’s rule does not contribute to asymmetric via-bility in centrarchids. First, there is no published evi-dence that reciprocal hybrid crosses produce differentsex ratios or that asymmetric viabilities are sex limited(Bolnick and Near 2005). Second, although biased sexratios are observed in some centrarchid hybrids, theycannot be explained by inviability of one sex, becausehybrid hatching success remains high (.90%) in manycrosses with strongly biased sex ratios (Bolnick andNear 2005). We hypothesize that sex-ratio bias mayreflect a breakdown in sex determination rather thanHaldane’s rule (Bolnick and Near 2005). Third, thereis no evidence that the centrarchid genome includes alarge hemizygous region linked to sex. Karyotypic stud-ies have identified no heteromorphic sex chromosomes(Roberts 1964). A genomic scan using AFLPs foundvery low levels of sex linkage (Lopez-Fernandez andBolnick 2007). If indeed the sex-linked region of thegenome is small, there may be little opportunity forhemizygous sex-linked loci to generate DMIs and asym-metric incompatibility (Turelli and Begun 1997) ½butthe ‘‘effective’’ size of the sex-linked region may be mag-nified by genetic conflicts over sex ratio (Tao et al.2007a,b)�.
In conclusion, our results provide the first tentativeevidence that asymmetric reproductive isolation is atleast partly due to systematic biases in molecular evolu-tionary rates. This suggests that mito-nuclear incompat-ibilities may have a general role in reproductive isolationamong centrarchid species, as asymmetries have beenfound in nearly all species pairs for which reciprocal-cross data are available. This general pattern of asym-metrical F1 viabilities, termed Darwin’s corollary byTurelli and Moyle (2007), remains relatively poorlyunderstood. Elaborating on these mechanisms mayhelp illuminate the mechanisms of speciation. Asym-
metric hybrid viabilities may also have important con-sequences for modern taxa, because hybrid viabilitiesmay influence the direction of introgression duringhybridization. For example, asymmetric mito-nuclearincompatibilities might contribute to the widespread ten-dency for mitochondrial DNA to introgress asymmetri-cally between hybridizing populations (Wilson andBernatchez 1998; Chan and Levin 2005).
We thank D. Rand and L. Moyle for comments. This research wassupported by National Science Foundation grants IBN-0076436 toP.C.W. and DEB-0412802 to D.I.B., the University of Texas at Austin,the University of California at Davis, and Yale University.
LITERATURE CITED
Barbash, D. A., D. F. Siino, A. M. Tarone and J. Roote, 2003 A rap-idly evolving MYB-related protein causes species isolation in Dro-sophila. Proc. Natl. Acad. Sci. USA 100: 5302–5307.
Bolnick, D. I., and T. J. Near, 2005 Tempo of hybrid inviability insunfish (Centrarchidae). Evolution 59: 1754–1767.
Bolnick, D. I., T. J. Near and P. C. Wainwright, 2006 Body sizedivergence promotes post-zygotic reproductive isolation in cen-trarchids. Evol. Ecol. Res. 8: 903–913.
Brideau, N. J., H. A. Flores, J. Wang, S. Maheshwari, X. Wang et al.,2006 Two Dobzhansky-Muller genes interact to cause hybridlethality in Drosophila. Science 314: 1292–1295.
Burton, R. S., C. K. Ellison and J. S. Harrison, 2006 The sorrystate of F2 hybrids: consequences of rapid mitochondrial DNAevolution in allopatric populations. Am. Nat. 168: S14–S24.
Chan, K. M. A., and S. A. Levin, 2005 Leaky prezygotic isolation andporous genomes: rapid introgression of maternally inheritedDNA. Evolution 59: 720–729.
Coyne, J. A., and H. A. Orr, 1989a Two rules of speciation, pp. 180–207 in Speciation and its Consequences, edited by D. Otte and J. A.Endler. Sinauer Associates, Sunderland, MA.
Coyne, J. A., and H. A. Orr, 1989b Patterns of speciation in Dro-sophila. Evolution 43: 362–381.
Coyne, J. A., and H. A. Orr, 2004 Speciation. Sinauer Associates,Sunderland, MA.
Darwin, C., 1859 On the Origin of Species. Harvard University Press,Cambridge, MA.
De Renzis, S., O. Elemento, S. Tavazoie and E. F. Wieschaus,2007 Unmasking activation of the zygotic genome using chro-mosomal deletions in the Drosophila embryo. PLoS Biol. 5:1036–1051.
Dettman, J. R., D. J. Jacobson, E. Turner, A. Pringle and W. J.Taylor, 2003 Reproductive isolation and phylogenetic diver-gence in Neurospora: comparing methods of species recognitionin a model eukaryote. Evolution 57: 2721–2741.
Dobzhansky, T., 1934 Studies on hybrid sterility. I. Spermatogene-sis in pure and hybrid Drosophila pseudoobscura. Proc. Natl. Acad.Sci. USA 19: 397–403.
Felsenstein, J., 1985 Phylogenies and the comparative method.Am. Nat. 125: 1–15.
Fitzpatrick, B. M., 2002 Molecular correlates of reproductive iso-lation. Evolution 56: 191–198.
Gallant, S. L., andD. J.Fairbain, 1997 Patternsofpostmatingrepro-ductive isolation in a newly discovered species pair, Aquarius remigisand A. remigoides (Hemiptera; Gerridae). Heredity 78: 571–577.
Haig, D., 2004 Genomic imprinting and kinship: How good is theevidence? Annu. Rev. Genet. 38: 553–585.
Haldane, J. B. S., 1922 Sex ratio and unisexual sterility in animalhybrids. J. Genet. 12: 101–109.
Harrison, R. G., 1983 Barriers to gene exchange between closelyrelated cricket species. I. Laboratory hybridization studies. Evolu-tion 37: 245–251.
Harrison, J. S., and R. S. Burton, 2006 Tracing hybrid incompat-ibilities to single amino acid substitutions. Mol. Biol. Evol. 23:559–564.
Kenyon, L., and T. Moraes Carlos, 1997 Expanding the functionalhuman mitochondrial DNA database by the establishment of pri-
Asymmetric Postmating Isolation in Centrarchid Fishes 1047
mate xenomitochondrial cybrids. Proc. Natl. Acad. Sci. USA 94:9131–9135.
Laurie, C. C., 1997 The weaker sex is heterogametic: 75 years ofHaldane’s rule. Genetics 147: 937–951.
Lopez-Fernandez, H., and D. I. Bolnick, 2007 What causes partialF1 hybrid viability? Incomplete penetrance versus genetic varia-tion. PLoS One 2:e1294.
Merriner, J. V., 1971 Development of intergeneric centrarchid hy-brid embryos. Trans. Am. Fish. Soc. 100: 611–618.
Muller, H. J., 1939 Reversibility in evolution considered fromthe standpoint of genetics. Biol. Rev. Camb. Philos. Soc. 14:261–280.
Muller, H. J., 1942 Isolating mechanisms, evolution and tempera-ture. Biol. Symp. 6: 71–125.
Navajas, M., A. Tsagkarov, J. Lagnel and M. J. Perrot-Minnot,2000 Genetic differentiation in Tetranychus urticae (Acari: Tetra-nychidae): Polymorphism, host races or sibling species? Exp.Appl. Acarol. 24: 365–376.
Near, T. J., D. I. Bolnick and P. C. Wainwright, 2004 Investigatingphylogenetic relationships of the Centrarchidae (Actinopterygii:Perciformes) using DNA sequences from mitochondrial and nu-clear genes. Mol. Phylogenet. Evol. 32: 344–357.
Near, T. J., D. I. Bolnick and P. C. Wainwright, 2005 Fossil cali-brations and molecular divergence time estimates in centrarchidfishes (Teleostei: Centrarchidae). Evolution 59: 1768–1782.
Noor, M. A. F., and J. L. Feder, 2006 Speciation genetics: evolvingapproaches. Nat. Rev. Genet. 7: 851–861.
Oliver, C. G., 1978 Experimental hybridization between the nym-phalid butterflies Phyciodes tharos and P. campestris montana. Evo-lution 32: 594–601.
Orr, H. A., 1993 A mathematical model of Haldane’s rule. Evolu-tion 47: 1606–1611.
Orr, H. A., 1995 The population genetics of speciation: the evolu-tion of hybrid incompatibilities. Genetics 139: 1805–1813.
Orr, H. A., and D. C. Presgraves, 2000 Speciation by postzygoticisolation: forces, genes, and molecules. BioEssays 22: 1085–1094.
Orr, H. A., and M. Turelli, 2001 The evolution of postzygotic iso-lation: accumulating Dobzhansky-Muller incompatibilities. Evo-lution 55: 1085–1094.
Philipp, D. P., H. R. Parker and G. S. Whitt, 1983 Evolution ofgene regulation: isozymic analysis of patterns of gene expressionduring hybrid fish development. Genet. Evol. 10: 193–237.
Posada, D., and K. A. Crandall, 1998 Modeltest: testing the modelof DNA substitution. Bioinformatics 14: 817–818.
Presgraves, D. C., 2002 Patterns of postzygotic isolation in Lepi-doptera. Evolution 56: 1168–1183.
Presgraves, D. C., 2003 A fine-scale genetic analysis of hybridincompatibilities in Drosophila. Genetics 163: 955–972.
Presgraves, D. C., and H. A. Orr, 1998 Haldane’s rule in taxa lack-ing hemizygous X. Science 282: 952–954.
Presgraves, D. C., L. Balagopalan, S. M. Abmayr and H. A. Orr,2003 Adaptive evolution drives divergence of a hybrid inviabil-ity gene between two species of Drosophila. Nature 423: 715–719.
Rakocinski, C. F., 1984 Aspects of reproductive isolation betweenCampostoma oligolepis and Campostoma anomalum pullum (Cyprini-formes: Cyprinidae) in northern Illinois. Am. Midland Nat. 112:138–145.
Rand, D. M., A. G. Clark and L. M. Kann, 2001 Sexually antagonis-tic cytonuclear fitness interactions in Drosophila melanogaster.Genetics 159: 173–187.
Rand, D. M., R. A. Haney and A. J. Fry, 2004 Cytonuclear coevolu-tion: the genomics of cooperation. Trends Ecol. Evol. 19: 645–653.
Rawson, P. D., and R. S. Burton, 2002 Functional coadaptation be-tween cytochrome c and cytochrome c oxidase within allopatricpopulations of a marine copepod. Proc. Natl. Acad. Sci. USA 99:12955–12958.
Roberts, F. L., 1964 A chromosome study of twenty species of Cen-trarchidae. J. Morphol. 115: 401–418.
Sackton, T. B., R. A. Haney and D. M. Rand, 2003 Cytonuclearcoadaptation in Drosophila: disruption of cytochrome c oxidaseactivity in backcross genotypes. Evolution 57: 2315–2325.
Schartl, M.,U.Hornung, H.Gutbrod, J.-N. VolffandJ.Whitbrodt,1999 Melanoma loss-of-function mutants in xiphophorus causedby Xmrk-oncogene deletion and gene disruption by a transposableelement. Genetics 153: 1385–1394.
Sturtevant, A. H., 1920 Genetic studies on Drosophila simulans. I.Introduction. Hybrids with Drosophila melanogaster. Genetics 5:488–500.
Swofford, D. L., 2002 PAUP*. Phylogenetic Analysis Using Parsimony(*and Other Methods), Version 4. Sinauer, Sunderland, MA.
Tao, Y., J. P. Masly, L. Araripe, Y. Ke and D. L. Hartl, 2007a A sex-ratio meiotic drive system in Drosophila simulans. I. An autosomalsuppressor. PLoS Biol. 5: e292.
Tao, Y., L. Araripe, S. B. Kingan, Y. Ke, H. Xiao et al., 2007b A sex-ratio meiotic drive system in Drosophila simulans. II. An X-linkeddistorter. PLoS Biol. 5:e293.
Thornton, W. A., 1955 Interspecific hybridization in Bufo woodhouseiand Bufo valliceps. Evolution 9: 455–468.
Tiffin, P., M. S. Olson and L. C. Moyle, 2001 Asymmetrical cross-ing barriers in angiosperms. Proc. R. Soc. Lond. 268: 861–867.
Ting, C.-T., S.-C. Tsaur, M.-L. Wu and C. I. Wu, 1998 A rapidly evolv-ing homeobox at the site of a hybrid sterility gene. Science 282:1501–1504.
Turelli, M., 1998 The causes of Haldane’s rule. Science 282: 889–891.
Turelli, M., and D. J. Begun, 1997 Haldane’s rule and X-chromo-some size in Drosophila. Genetics 147: 1799–1815.
Turelli, M., and L. C. Moyle, 2007 Asymmetric postmating isola-tion: Darwin’s corollary to Haldane’s rule. Genetics 176: 1059–1088.
Turelli, M., and H. A. Orr, 1995 The dominance theory of Haldane’srule. Genetics 140: 389–402.
Turelli, M., and H. A. Orr, 2000 Dominance, epistasis, and thegenetics of postzygotic isolation. Genetics 154: 1663–1679.
Welch, J. J., 2004 Accumulating Dobzhansky-Muller incompatibili-ties: reconciling theory and data. Evolution 58: 1145–1156.
West, J. L., and F. E. Hester, 1966 Intergeneric hybridization ofcentrarchids. Trans. Am. Fish. Soc. 95: 280–288.
Whitt, G. S., D. P. Philipp and W. F. Childers, 1977 Aberrant geneexpression during the development of hybrid sunfishes (Perci-formes, Teleosti). Differentiation 9: 97–109.
Willett, C. S., and R. S. Burton, 2001 Viability of cytochrome Cgenotypes depends on cytoplasmic backgrounds in Tigriopus cal-ifornicus. Evolution 55: 1592–1599.
Wilson, C. C., and L. Bernatchez, 1998 The ghost of hybrids past:fixation of arctic charr (Salvelinus alpinus) mitochondrial DNA inan introgressed population of lake trout (S. namaycush). Mol.Ecol. 7: 127–132.
Wolfram, S., 2003 The Mathematica Book, Ed. 5. Wolfram Media,Champaign, IL.
Wu, C.-I., and A. W. Davis, 1983 Evolution of post-mating reproduc-tive isolation - the composite nature of Haldane’s rule and its ge-netic bases. Am. Nat. 142: 187–212.
Communicating editor: L. Harshman
1048 D. I. Bolnick et al.